Six capability areas — from first-principles physics through to a native desktop app and open REST API. Every layer is built on the same validated simulation engine.
15 independent sub-models. No curve-fitting. Every coefficient traceable to a physical measurement.
MotoQuant's physics engine integrates 15 coupled differential equations at a 1 ms timestep using a fixed-step RK4 integrator (with an optional adaptive solver). Unlike spreadsheet calculators that treat power-to-weight ratio as the only variable, MotoQuant models:
Pacejka tire model with thermal evolution, so grip changes as the tyre heats. Aerodynamic drag with a per-bike Cd×A (0.308–0.68 m²) — no two bikes share the same number. Gear shifting and clutch engagement dynamics including slipper clutch behaviour. Forced induction — turbo lag, supercharger mechanical loss, nitrous injection windows. Engine inertia and parasitic losses (alternator, oil pump, water pump, valvetrain).
The adaptive-dt solver is especially useful for identifying the precise clutch engagement window that produces the minimum 60-ft time. Reference baselines — GSX-R 1000 K5 at 10.033s, R1 at 10.286s, Hayabusa Gen1 at 10.471s — are all within 0.001s of published Cycle World and Sport Rider timeslips.
Screenshot coming soon
424 parts. Cost-per-tenth ranking. Budget-constrained knapsack optimiser.
The parts catalog covers 465 upgrades across 20 categories — exhausts, ECU flashes, forced induction kits, slipper clutches, suspension, brakes, wheelie bars. Every part has a physics-accurate parameter delta: an Akrapovič slip-on doesn't add a round number of horsepower, it shifts the entire torque curve.
The ROI engine runs a 0/1 knapsack DP at ₹100 bucket resolution across any compatible parts set. Tell it your budget and it finds the combination that minimises your ET. The cost-per-tenth column tells you how many rupees you're paying per 0.1 second of ET gain — the clearest way to compare a ₹15,000 exhaust against a ₹60,000 turbo kit.
Build profiles — Street Legal, Drag Prep, Full Race — pre-filter the catalog so you don't accidentally spec a nitrous kit onto a street-registered bike.
Screenshot coming soon
Dragy GPS, dyno sheets, ECU logs. Auto-calibrate Cd and μ to match your actual strip times.
Physics is only as good as its inputs. MotoQuant accepts three data sources:
Dragy GPS logs (CSV, JSON, or RaceChrono VBO-derived exports) — the importer detects run boundaries, finds the staged position, integrates v(t) to distance, and interpolates ET at exactly 402.336 m with sub-sample accuracy. The best-run selector removes outliers from multi-run files.
Dyno sheets (DynoJet CSV, generic metric CSV, MoTeC tab-delimited) with full SAE J1349 atmospheric correction so power figures are apples-to-apples regardless of dyno location.
ECU logs (Woolich Racing, Power Commander PC-Link, MoTeC i2) with automatic channel mapping and run-window detection from the speed channel.
Once you have a real run imported, the auto-calibrator fits Cd + μ_peak jointly using differential evolution (global) followed by Nelder-Mead refinement (local). The result is a BikeConfig that matches your specific bike, rider, and track surface — not a generic factory spec.
Screenshot coming soon
BoTorch Gaussian-process optimizer. 7-parameter continuous search. Budget-aware.
Beyond the parts catalog, MotoQuant ships a full optimization stack:
Sobol global sensitivity analysis quantifies which parameters matter most for your specific bike. On a stock Hayabusa, aero Cd and μ_peak have similar sensitivity. On a built Pulsar NS200, the clutch engagement RPM dominates.
Bayesian optimization (BoTorch) searches a 7-parameter continuous space — power, mass, Cd, frontal area, tire μ, tire pressure, final drive ratio — within a budget constraint. It converges in 30–50 iterations using expected improvement acquisition. No grid search required.
The Smart Recommendations engine wraps the knapsack with synergy detection: if you've selected an exhaust, it weights ECU flashes higher because the pair delivers super-linear gains. Compatibility rules block physically nonsensical combinations (dual exhausts, stacked ECUs).
Natural language queries — "optimize my Hayabusa for 35°C at Aamby Valley, budget ₹50k" — resolve against the live catalog and return a full recommendation set.
Screenshot coming soon
Tauri 2 + React 19. Python physics sidecar. Offline-first. No browser required.
The MotoQuant desktop app (currently in closed beta) is a Tauri 2 shell wrapping a React 19 frontend. It spawns a PyInstaller-packaged FastAPI sidecar on an ephemeral port at startup — so you get the full 15-sub-model physics engine locally, with no cloud dependency.
Five pages ship in the current build: Simulate (bike + venue + sliders + 4-panel chart — speed, RPM, g-force, power), Bikes (searchable paginated catalog), Build (parts composer with live ET delta on every toggle), Compare (A-vs-B head-to-head with overlaid v(t) chart), and History (full leaderboard from local DB).
The physics sidecar is the same code that powers the web API — same importer, same gear-ratio library, same 15 sub-models. Results are deterministic across web and desktop.
Desktop beta access is available to tuning shops and racing teams. Contact motoquant@gmail.com to apply.
Screenshot coming soon
FastAPI. Full timeseries. CORS-open for web integrations.
Every feature the web UI uses is accessible via the public REST API hosted at web-production-de6eb.up.railway.app.
POST /api/simulate — full simulation returning complete v(t), RPM, gear, acceleration, and ET. Typical latency 80–150 ms. POST /api/simulate/quick — ET + trap speed + splits only, ~50 ms. Used by the slider-driven BuildPage. POST /api/simulate/compare — A-vs-B comparison, returns both timeseries with delta ET/trap and cost-to-tenth. POST /api/calibrate/dragy — upload a Dragy CSV and get run-boundary detection + best-run ET/trap/splits, with optional sim overlay. POST /api/calibrate/dyno — upload a dyno CSV and get the corrected torque curve ready for injection. GET /api/bikes — paginated catalog with manufacturer/category filters. GET /api/venues — all 20 venues with 12-month seasonal weather data.
Interactive docs at /docs (FastAPI's automatic OpenAPI UI). No authentication for basic simulation — rate-limited to 60 requests/minute per IP. Authenticated API keys (email to request) unlock higher limits and private run storage.
Screenshot coming soon
Free in the browser. No account required to start.